In enterprise networking, failure is not an option. Wired networks have operated under this reality for a couple of generations, but wireless LANs have been a step or two behind. Yet now that many enterprises are adopting wireless as the primary network access layer,
WLAN controllers: Single point of failure
Every network architect knows that a WLAN controller is a single point of failure. That's why many mission-critical enterprise networks have controllers deployed in high-availability (HA) pairs that stay synced so the production controller can fail over quickly and access point (AP) downtime is minimized.
Controllers, however, are expensive. Pairing each of those boxes with a dedicated backup will bust a budget, so many companies deploy controllers in an n+1 configuration, where one large backup controller sits in standby mode ready to serve as a failover backup for multiple production controllers. N+1 saves money, but it also delays failover because it's impossible for the backup controller to sync itself without multiple controllers. Instead, the backup controller has to rebuild the profile for the downed controller, adding minutes of downtime. In the meantime, clients on the APs are left hanging. Phone calls drop. Videos freeze. Email stops working. Users flood the help desk with tickets.
More on WLAN controllers
Cloud-based controllers offer WLAN survivability in branch offices
Wither WLAN controllers? Fat APs making a comeback, sort of
"I've heard of some failover times that last 10 minutes. That's totally unacceptable," said Andre Kindness, senior analyst with Forrester Research Inc. "Mostly it's a problem of vendor architectures. All these guys have the control plane in one location. If it fails, how do you move all that information over [to a backup controller] and get it back up and running?"
Controllers require a lot of sophisticated software to move from one location to another. "In a one-for-one, you can share a heartbeat back and forth [so that failover last seconds or a couple of minutes]. But that's too expensive, so most companies go with n+1. That's where you get 10 minutes," Kindness explained.
Aruba customer Tony Alphier, IT director at Regional Medical Center at Memphis, has a single HA pair of controllers managing his wireless LAN, so he expects failover to happen quickly.
"We're the level one trauma facility in the region, and we have a number of applications we've developed on the iPad," Alphier said. "When the trauma physicians do their rounds, they rely on those applications and the wireless signal. Having stable wireless is critical."
When Alphier first deployed his Aruba network last year, he performed a failover test during the switchover from his legacy Cisco network. "It was able to handle it with minimal downtime. It took two to three minutes, and then all the APs came back up. There weren't too many issues surrounding connectivity. Some people lost it, but by and large it worked out very well."
Making controller failover and AP recovery faster
Wireless LAN resiliency isn't necessarily an arms race, but vendors are constantly trying to improve the state of the art in this area. "The functionality required for fast failover has been required for mission-critical IT organizations for a while," said Tim Zimmerman, research vice president at Gartner. All vendors are striving for fast, if not instant, failover to serve the growing number of IT organizations that rely on wireless LANs for primary network access, he said.
In an incremental update to its AruabaOS software released this week, Aruba announced a 50% to 60% improvement, to two or three seconds, in failover time, even in n+1 configurations. Aruba also announced enhancements to ensure that rich media applications like FaceTime and Lync don't drop sessions during failover.
"Now [Aruba] is saying [failover] is in the milliseconds-to-second range, where it's pretty much transparent to the end user," Alphier said. "If you lost connectivity for five or ten seconds, the application doesn't stop working."
Recovery time is not zero, "but it almost feels like a roaming event," said Ozer Dondurmacioglu, Aruba product marketing manager. Aruba has enhanced its software for client devices and applications so that voice and video communications only pause rather than drop during failover, he said. No manual intervention or re-authentication by the end user is needed.
Meanwhile, Cisco will cut its failover time in its architecture with a wireless LAN software release that comes out next week that introduces a new feature: AP Stateful Switch Over (SSO). In wireless LANs with one-to-one, HA pairs of controllers, actual AP downtime will drop to 300 milliseconds. "Also, the keys are cached between the primary and secondary controller, so the client doesn't' have to re-authenticate with the Radius server, which is a bottleneck," said Greg Beach, director of product marketing at Cisco.
Early next year Cisco will introduce a Client SSO feature, which will eliminate the need for clients to re-authenticate at all after a controller failover.
Make backup controllers cheaper
Another solution to reducing failover time is to pair each production controller with a backup, but as mentioned above, that's expensive. Wireless LAN vendors Bluesocket [acquired by Adtran] and Meru Networks have introduced virtual WLAN controllers, which offer some savings. Cisco, known for charging customers a premium price for its hardware, is switching things up. It will now offer discounted SKUs for controllers that are deployed as HA backups.
"In the past, the price for a redundant controller was the same as an active controller," Beach said. "One Wireless Services Module for a Catalyst 6500 in the past cost $168,000 for 1,000 access points in a one-to-one configuration. Now the second module will be priced at $25,000. The 5500 controller with a maximum of 500 access points is $105,000. Now the high availability version will be $20,000."
Cisco did not offer a full price list for HA controllers, but all of them would be significantly lower than SKUs for production boxes, Beach said.
Let us know what you think about the story; email Shamus McGillicuddy, News Director.